On Sat, Jan 23, 2021 at 9:29 AM Drew Fustini <drew@xxxxxxxxxxxxxxx> wrote: > > This RFC is a change in approach from my previous RFC patch [1]. It adds > "pinnux-select" to debugfs. Function and group on the pin control device > will be activated when 2 integers "<function-selector> <group-selector>" > are written to the file. The debugfs write operation pinmux_select() > handles this by calling ops->set_mux() with fsel and gsel. ... > RFC notes: Please, move below to reST formatted document. ... > +static ssize_t pinmux_select(struct file *file, const char __user *user_buf, > + size_t cnt, loff_t *ppos) > +{ > + struct seq_file *sfile = file->private_data; > + struct pinctrl_dev *pctldev = sfile->private; > + const struct pinmux_ops *ops = pctldev->desc->pmxops; > + int fsel, gsel, ret; > + // RFC note: two integers separated by a space should never exceed 16 > + char buf[16]; > + if (*ppos != 0) > + return -EINVAL; But why? Do we really care about it? Moreover, you have no_llseek() below. > + ret = strncpy_from_user(buf, user_buf, cnt); Potential buffer overflow. cnt -> sizeof(buf) > + if (ret < 0) > + return ret; > + buf[cnt] = '\0'; Not sure, shouldn't be buf[sizeof(buf) - 1] = '\0'; ? > + if (buf[cnt - 1] == '\n') > + buf[cnt - 1] = '\0'; strstrip() ? > + ret = sscanf(buf, "%d %d", &fsel, &gsel); > + if (ret != 2) { > + dev_err(pctldev->dev, "%s: sscanf() expects '<fsel> <gsel>'", __func__); __func__ is useless, please drop it. And below as well. > + return -EINVAL; > + } > + ret = ops->set_mux(pctldev, fsel, gsel); > + if (ret != 0) { I thought I gave you a comment on this... if (ret) > + dev_err(pctldev->dev, "%s(): set_mux() failed: %d", __func__, ret); > + return -EINVAL; > + } > + > + return cnt; > +} ... > debugfs_create_file("pinmux-pins", S_IFREG | S_IRUGO, > devroot, pctldev, &pinmux_pins_fops); > + debugfs_create_file("pinmux-select", 0200, > + devroot, pctldev, &pinmux_set_ops); Consider to add another (prerequisite) patch to get rid of symbolic permissions. -- With Best Regards, Andy Shevchenko